An Efficient Data-Driven Routing Protocol for Wireless Sensor ...

2 downloads 1411 Views 323KB Size Report
protocol (called DDRP) for wireless sensor networks with mobile .... 4 State key Lab of Networking & Switching Tech., Beijing University of Posts and ...
This full text paper was peer reviewed at the direction of IEEE Communications Society subject matter experts for publication in the IEEE ICC 2011 proceedings

An Efficient Data-Driven Routing Protocol for Wireless Sensor Networks with Mobile Sinks Lei Shi 1,2, Baoxian Zhang 1, Kui Huang 3, Jian Ma 4 1

Research Center of Ubiquitous Sensor Networks, Graduate University of Chinese Academy of Sciences, Beijing 100049, China 2 College of information & Electronic Engineering, Zhejiang Gongshang University, Hangzhou, Zhejiang 310018, China 3 Research Center of Internet of Things, Institute of Microelectronics, Chinese Academy of Sciences, Beijing 100029, China 4 State key Lab of Networking & Switching Tech., Beijing University of Posts and Telecommunications, Beijing 100876, China

Abstract—In this paper, we propose a data-driven routing protocol (called DDRP) for wireless sensor networks with mobile sinks (mWSNs). The design objective of DDRP is to effectively reduce the protocol overhead for data gathering in such networks. DDRP exploits the broadcast feature of wireless transmissions for sensor nodes for (gratuitous) route learning. To achieve this goal, each data packet carries an additional option recording the known distance from the sender of the packet to the destined mobile sink. The overhearing of such a data packet will gratuitously provide listeners a route to mobile sink. This is the so-called data-driven nature of DDRP. Continuous such route-learning among neighboring nodes will provide route information for more and more sensor nodes in the network. We present the detailed design of the DDRP protocol. Simulation results show that DDRP has much lower protocol overhead as compared with existing work while ensuring high packet delivery ratio. Keywords-routing; protocol overhead; mobile sink; wireless sensor network

I. INTRODUCTION Recently, introducing mobile sinks into Wireless Sensor Networks (WSN) has attracted a lot of attention. This leads to several advantages such as hotspot removal, longer network lifetime, and energy use optimizations. However, sink mobility can cause unexpected changes of network topology and it may bring excessive protocol overhead for route discovery and maintenance. The excessive protocol overhead might offset the benefits from using mobile sinks. Therefore, the performance of a wireless sensor network with mobile sinks (mWSN) highly depends on how the routing protocol can be designed to keep high data delivery ratio while suppressing the protocol overhead caused by sink mobility. In this paper, we propose a data-driven routing protocol (DDRP) for mWSNs. The design objective is to effectively reduce the protocol overhead for data gathering in such networks. To achieve this goal, DDRP integrates data-driven packet forwarding and random walk. DDRP exploits the broadcast feature of wireless transmissions for gratuitous route learning. For this purpose, each data packet needs to carry an additional option recording the known distance from the sender to the target mobile sink. The overhearing of such a data packet This work was supported in part by National Natural Science Foundation of China under Grant Nos. 60970137 and 61070166, National Key Special Program of China under Grant No. 2010ZX03006-001-02.

will provide the listening nodes a route to mobile sink. This is the so-called data-driven nature of DDRP. Continuous such route-learning (starting from the adjacent sensor nodes of mobile sinks) will provide route information for more and more sensor nodes in the network. When a sensor node has no valid route for reaching a mobile sink, random walk is used until the data packet reaches a sensor node with a route to mobile sink or a sink directly. DDRP is very simple to implement and introduces little extra control overhead for route learning. Simulation results demonstrate that DDRP can achieve significant reduction in terms of protocol overhead under different scenarios while ensuring high packet delivery ratio. II. RELATED WORK Existing work on design of routing protocols for mWSNs can be divided into two types. The first type is to use mobile collectors for data gathering. A sensor node only reports its sensed data when a mobile collector is in its communication range. The other type is to use mobile sinks such that a sensor can report its sensed data via a multi-hop path. Protocols falling into the category of using mobile collectors include data mules [1] and mobile collectors with mobility predictions [2]. In [1], data mules are used for collecting data from sparser sensor networks and targeted for delay-tolerant applications. In [2], Lee et al. propose a Predictable Mobility-based Data Dissemination protocol (PMDD). Based on the predicted location and time of mobile collector, a sensor can send its data towards a location on sink’s moving path without frequent sink location updates. In general, data gathering protocols using the above mobile collectors cause little communication overhead, but often result in long delivery latency (e.g., a few hours or even days) and large storage overhead at sensor nodes. Protocols falling into the category of WSNs with mobile sinks encourage sensors in the network to report their data via multihop paths. The routing paths can be established with the assistance of location information or via the network wide flooding of an “interest” packet for gradient establishment. This type of protocols can gather sensing data in a timely fashion (e.g., in a few hundreds of milliseconds). In this paper, our routing protocols belonging to this type. Next, we will review existing work in this area. TTDD (Two-Tier Data Dissemination) [3] provides scalable

978-1-61284-231-8/11/$26.00 ©2011 IEEE

This full text paper was peer reviewed at the direction of IEEE Communications Society subject matter experts for publication in the IEEE ICC 2011 proceedings

and efficient data delivery from a data source to multiple mobile sinks. Each data source pro-actively builds a grid structure by dividing the sensor field into cells with dissemination nodes located at the crossing points of the grid. The delivery structure by TTDD is easy to maintain. However, with the increase of data source number and sink number, TTDD can produce excessive protocol overhead. In [4], Wang et al. designed a Local Update-based Routing Protocol (LURP) to decrease energy consumption for mWSNs caused by frequent location updates. In [5], Tian et al. proposed an Anchor-based Voronoi Routing Protocol (AVRP) and a Trail-based Forwarding Routing Protocol (TRAIL), respectively, for mWSNs. AVRP introduces an anchor (the closest neighbor sensor of mobile sink) based on a delivery structure by dynamic Voronoi scoping in order to suppress the overall updating overhead. However, AVRP still needs to flood interest packets for building delivery structure as sinks move, which could cause the scalability issue. In TRAIL, each sensor records the most recent instant as mobile sink trail information when a mobile sink passes by it. To enable packet forwarding at a packet holder along a sink trail, an on-demand query-reply cycle is executed to identify the most promising direction for further forwarding a data packet along the trail. However, a sink trail can be fragile in sparse mWSNs and can be outdated as sinks move. In [6][7], the authors mainly focused on how to control the mobility of sinks in order to effectively prolong the network lifetime, which is not the focus of this paper. III. PROTOCOL DESIGN OF DDRP In this section, we will first give an overview regarding how DDRP works and then present its detailed design description. A. Protocol Overview In DDRP, we assume that sensor nodes are homogeneous such that all sensor nodes and mobile sinks have the same communication range. Moreover, no location information of nodes is assumed. Each mobile sink broadcasts beacon messages periodically to its one-hop neighboring sensor nodes as it moves so that each neighboring sensor node of a mobile sink can easily learn the fact that it can report data to a sink directly. A beacon message carries the following information: the ID of the mobile sink, a time stamp, and beacon interval, whose length can be affected by the velocity of the mobile sink and nodes’ communications range and thus needs to be tuned based on specific network scenarios. Beacon messages will not be propagated further. Each data packet carries an option, Dist2mSink, which records the shortest known distance from the sender of the packet to a mobile sink. For instance, when Dist2mSink equals to two, it means that the sender of the packet is currently two-hop away from a mobile sink. If the maximum allowed value of Dist2mSink is set as K (K≥1), the number of bits required for storing Dist2mSink is thus ⎡log2(K+1)⎤. This means the option contains at most K+1 values. If Dist2mSink equals to K+1 or larger, it means that the sensor node has no knowledge about any mobile sink and thus K+1 means

INFINITY. Sensor nodes in the network is divided into three types according to their hop distance away from their respective nearest mobile sink(s): One-hop neighboring sensor nodes of mobile sinks (OHS), Multihop neighboring sensor nodes of mobile sinks (MHS), and Infinite-hop neighboring sensor nodes of mobile sinks (IHS), which means the sensor nodes have no route to reach a mobile sink. Different types of sensors will have different procedures for route learning. Fig.1 illustrates how sensors may transit among different types. DDRP integrates data-driven forwarding and random walk and the key idea is as follows. Initially, all sensor nodes are IHS nodes. As a mobile sink moves, it periodically broadcasts beacon packets to its one-hop neighboring sensor nodes. Those sensor nodes that receive such beacon packets become OHS sensor nodes. Accordingly, each OHS sensor sets its Dist2mSink value as one and includes this in the data packets (if any), which it sends in the near future. When a potential two-hop MHS node overhears the data transmission by an OHS node, the MHS node updates the local routing table by using the Dist2mSink value carried in the received data packet. When the two-hop MHS node sends data packets, it sets the Dist2mSink as two and includes it in the data packets, then its three-hop MHS neighboring nodes will learn the existence of such a three-hop path to reach a mobile sink. Other nodes will operate similarly for route learning. This process continues until Dist2mSink=K+1, which is treated as INFINITY in DDRP. For a sensor node with a route to a mobile sink, it will forward its data packets along the known route. Each of such data packets will carry the option indicating the hop distance to the target sink (denoted by d, d